GraphTypes - Inworld AI Documentation

GraphTypes are specialized data types used for input and output between nodes in the Inworld Node.js Runtime graph system. These types provide structured data handling for various AI processing workflows including LLM interactions, audio processing, tool calls, and more.

Import

import { GraphTypes } from '@inworld/runtime/graph';

// Usage examples
const audioData = new GraphTypes.Audio({ /* audio data */ });
const content = new GraphTypes.Content({ /* content data */ });

Available Types

Name	Description	Common Usage
Audio	Audio data type for speech processing operations.	- Input for RemoteSTTNode - Output from RemoteTTSNode - Audio data handling in custom nodes
ClassificationResult	Classification analysis results from text or content analysis operations.	- Output from classification nodes - Input for classification-based routing
Content	Encapsulates LLM response content with optional tool calls. The primary output type from LLM chat nodes.	- Output from RemoteLLMChatNode - Input for processing LLM responses
ContentStream	Streaming content responses from LLM nodes, providing real-time text generation.	- Output from streaming LLM operations - Input for real-time response processing
Custom	Generic data type for custom node implementations.	- Input/output for custom nodes - Flexible container for application-specific data - Used when standard types don’t fit the use case
GoalAdvancement	Goal tracking and advancement data for monitoring progress toward objectives.	- Output from goal tracking nodes - Input for progress-based routing
KnowledgeRecords	Knowledge base query results containing retrieved information and metadata.	- Output from KnowledgeNode - Input for processing knowledge base responses
ListToolsResponse	Response containing available tools from MCP (Model Context Protocol) servers.	- Output from MCPListToolsNode - Input for tool selection logic
LLMChatRequest	Chat request data type for LLM processing, containing conversation history, tools, and configuration.	- Input for RemoteLLMChatNode - Built by LLMChatRequestBuilderNode
MatchedIntents	Intent detection results from natural language understanding.	- Output from intent classification nodes - Input for conditional routing based on detected intents
MatchedKeywords	Keyword matching results from text analysis.	- Output from KeywordMatcherNode - Input for keyword-based routing
MemoryState	Memory and conversation state information for maintaining context across interactions.	- Input/output for memory management nodes - Used for stateful conversation handling
TextChunks	Output from text chunking operations, containing segmented text data.	- Output from TextChunkingNode - Input for processing segmented text
TextStream	Streaming text data for continuous text processing operations.	- Input/output for text transformation nodes - Used in streaming text processing pipelines
ToolCallRequest	Tool execution request containing function calls and parameters.	- Output from tool call parsing nodes - Input for MCPCallToolNode
ToolCallResponse	Tool execution response containing results and status information.	- Output from MCPCallToolNode - Input for processing tool execution results
TTSRequest	Text-to-speech request containing text content and voice configuration.	- Input for RemoteTTSNode - Supports both direct text and streaming input
TTSOutputStream	Streaming audio output from TTS processing operations.	- Output from streaming TTS operations - Used in real-time audio generation scenarios

​Import

​Available Types

Import

Available Types